Novel Leakage Detection and Localization Method Based on Line Spectrum Pair and Cubic Interpolation Search

Abstract

Leakage detection in water distribution systems (WDS) is critical to ensuring the security of urban water supplies. Acoustic detection methods have been used for leakage detection in water utilities, but ambient noise in real cases interferes with their detection accuracy, and the localization process in meshed or looped pipe networks requires significant computational costs. To increase the effectiveness of acoustic detection methods in practical applications, the current work proposes a novel leakage detection and localization method. This method extracts line spectrum pairs (LSP) of leakage signals and uses a random forest (RF) model to detect leaks; then, a cubic interpolation search (CIS) algorithm is developed to locate leaks. The LSP-based leakage detection method shows a clear advantage over the detection methods based on linear prediction coefficients (LPC) and time or frequency domain features. The proposed leakage detection method achieves 99.45% accuracy. In the case of −5 dB, the detection accuracy reaches 93.89%. The CIS algorithm is found to be more stable and shows a faster convergence speed than a commonly used graph-based search algorithm. The localization error is low (i.e., 2.22 m to 9.99 m). The LSP-CIS combined algorithm provides a more effective solution for leakage detection and localization.